Policy discussion
The text of this post is on X; follow the source link to read it.
Jul 2, 2026 · Posts on X (state legislators) · policy discussion; Housing and Community Development · Source
97%
of State Senate Democrats score to the right of Chang on the CSIP scale
CSIP score −1.23 (approximate range −1.29 to −1.16, which understates the uncertainty); negative is left of the scale’s centre, positive right. From 656 donors and no roll calls.
CSIP places every official on one left–right scale estimated from roll-call votes, campaign donations and the text of statements; 0 is the centre of the scale, not a definition of moderate. The share says where Chang sits among State Senate Democrats, not how far from the centre.
At a glance
5 measures against State Senate Democrats. The policy areas Chang talks about and legislates on agree at 0.33 on a 0–1 scale, in the bottom 5% of State Senate Democrats. 47% of the 2,308 bills Chang sponsored got out of committee, in the bottom half of the Hawaii Senate.
The measures on which Chang sits furthest from the middle of State Senate Democrats, each against the same group, with the CSIP score when there is one. Every number is explained in its own section below.
CSIP ideology
−1.23
More liberal than 97% of State Senate Democrats · ± 0.06
Talk-to-bills agreement
0.33
State Senate Democrats average 0.56 · ranks 221st of 228
Out of committee
47%
Hawaii Senate average 50% · ranks 17th of 24
From Hawaii, 2026
92%
Of itemized individual dollars · ranks 393rd of 714 State Senate Democrats
Itemized, 2026
$68.7K
State Senate Democrats’ median $55.5K · ranks 341st of 746
Statements
47
47 posts on X
Peer group: State Senate Democrats. Rank 1 is the highest value in the group; the thick bar is Chang, the thin grey bar the peer average. The headline is the measure on which Chang is most unusual among peers (Methods, §08).
What Chang says
Chang attacks opponents personally less than 75% of State Senate Democrats. Personal attacks in 0% of 47 statements, against 2.4% for the average State Senate Democrat; 248th of 329.
An AI model reads each of Chang’s 47 statements in the HI 2025–26 session (47 posts on X) and marks the kinds of statement it contains: a personal attack, criticism of a policy, discussion of policy, bipartisanship, credit claiming, and extreme labels. A statement can carry several. Every rate is compared with the average for State Senate Democrats and ranked among them.
Kinds of statement
Share of statements carrying each label, the HI 2025–26 session
0%80%
Chang
Peers
Rank
Policy discussion26 statements
55%
43%
101st of 329
Policy criticism5 statements
11%
10%
130th of 329
Credit claiming4 statements
9%
16%
216th of 329
Bipartisanship1 statements
2%
3%
179th of 329
Personal attacks0 statements
0%
2%
248th of 329
Show the data
| Row | Chang | Peers |
|---|---|---|
| Policy discussion | 55% | 43% |
| Policy criticism | 11% | 10% |
| Credit claiming | 9% | 16% |
| Bipartisanship | 2% | 3% |
| Personal attacks | 0% | 2% |
47
statements classified
47 posts on X.
Show the data
| Part | Value |
|---|---|
| posts on X | 47 |
How to read the bars
The thick bar is Chang; the thin grey bar is the average State Senate Democrats. A longer thick bar means Chang does this more than peers. Rank 1 is the highest rate in the group.
Extreme labels
Chang used no extreme labels in 47 statements. The peer average is 0.8%.
No extreme labels in this window.
Who gets the labels
Targets named alongside an extreme label, by number of statements
No targets recorded.
What an extreme label is
Calls a politician, a party or a policy “far left”, “far right”, “MAGA”, “extreme”, “radical”, “woke”, “fascist” or “communist”. The target is who or what the label was aimed at. The lists show the five most used labels and targets; the toggle shows the rest.
Month by month
Share of statements per month, the HI 2025–26 session
Show the data
| Column | Statements | Personal attacks | Policy criticism | Policy discussion | Bipartisanship | Credit claiming | Extreme labels |
|---|---|---|---|---|---|---|---|
| Jan ’25 | 11 | 0.0% | 9.1% | 81.8% | 0.0% | 18.2% | 0.0% |
| Feb ’25 | 12 | 0.0% | 8.3% | 16.7% | 0.0% | 0.0% | 0.0% |
| Mar ’25 | 3 | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) |
| Apr ’25 | 1 | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) |
| May ’25 | 2 | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) |
| Jun ’25 | 2 | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) |
| Jul ’25 | 1 | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) |
| Aug ’25 | 0 | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) |
| Sep ’25 | 0 | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) |
| Oct ’25 | 0 | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) |
| Nov ’25 | 0 | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) |
| Dec ’25 | 0 | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) |
| Jan ’26 | 0 | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) |
| Feb ’26 | 0 | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) |
| Mar ’26 | 2 | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) |
| Apr ’26 | 0 | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) |
| May ’26 | 0 | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) |
| Jun ’26 | 0 | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) |
| Jul ’26 | 13 | 0.0% | 7.7% | 76.9% | 7.7% | 0.0% | 0.0% |
| Aug ’26 | 0 | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) |
| Sep ’26 | 0 | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) |
| Oct ’26 | 0 | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) | — (fewer than 10 statements) |
How to read the strips
Each row is one kind of statement; each cell is a month, shaded against that row’s own peak (darker = closer to the peak), with the peak and the latest value written at the right. The top row counts statements. The strip covers every month of the HI 2025–26 session, the same span as the rates above, including months with no statements. A month with fewer than 10 statements is hatched, not shaded: one or two posts would swing its rate to 0% or 100%. 19 months are hatched here; 13 of them have no statements.
In Chang’s words
Real statements from the classifier’s example set, with the labels it assigned
Bipartisanship
The text of this post is on X; follow the source link to read it.
Jul 2, 2026 · Posts on X (state legislators) · policy discussion; bipartisanship; Economics and Public Finance · Source
Credit claiming
The text of this post is on X; follow the source link to read it.
Mar 14, 2025 · Posts on X (state legislators) · policy discussion; credit claiming; Government Operations and Politics · Source
Examples by category
Up to five recent statements in each category
Overtly questions a named person’s character, integrity or loyalty, or calls them names. Criticising a record or a policy does not count.
No statements in this category in this window.
All 47 statements
Newest first, the HI 2025–26 session
The full feed
Five statements at a time, newest first; filters apply to the statements loaded so far. Export saves the statements that match. Sources: 47 posts on X.
Policy
Housing and Community Development is what Chang talks about most: 42% of policy statements, 6.3 times the peer share. Talk and legislation agree at 0.33 on a 0–1 scale (State Senate Democrats average 0.56; 221st of 228).
Each policy statement is tagged by the AI model with one or more of 25 policy areas from the Congressional Research Service list, so the shares say what Chang talks about. The bills Chang sponsors and cosponsors carry the same areas, so the two can be compared.
What Chang talks about
Share of policy statements by area, the HI 2025–26 session
0%47%
Chang
Peers
Housing and Community Development11 statements
42%
7%
Economics and Public Finance6 statements
23%
10%
Education3 statements
12%
15%
Government Operations and Politics3 statements
12%
14%
Families2 statements
8%
5%
Show the data
| Row | Chang | Peers |
|---|---|---|
| Housing and Community Development | 42% | 7% |
| Economics and Public Finance | 23% | 10% |
| Education | 12% | 15% |
| Government Operations and Politics | 12% | 14% |
| Families | 8% | 5% |
How to read the bars
The thick bar is the share of Chang’s policy statements in an area; the thin grey bar is the average share for State Senate Democrats. A statement counts once in each of its areas, so shares can add to more than 100%. The five largest areas are shown; the toggle shows the rest.
Talks about · legislates on
Left: share of statements by policy area. Right: share of bills. Areas in order of their share of statements; both sides run from 0 to 47%. Grey ticks are the State Senate Democrats averages.
Show the data
| Area | Talks about | State Senate Democrats, talks about | Legislates on | State Senate Democrats, legislates on |
|---|---|---|---|---|
| Housing and Community Development | 42% | 7% | 9% | 5% |
| Economics and Public Finance | 23% | 10% | 2% | 3% |
| Education | 12% | 15% | 5% | 9% |
| Government Operations and Politics | 12% | 14% | 9% | 8% |
| Families | 8% | 5% | 1% | 3% |
0.33
talk-to-bills agreement, 0 to 1
State Senate Democrats average 0.56. Ranks 221st of 228.
What agreement means, and what it doesn’t
Agreement is 1 minus the distance between the two distributions (Jensen–Shannon): 1 means Chang legislates on exactly the areas Chang talks about, in the same proportions; 0 means no overlap. It does not say whether the bills match the positions in the statements, only the subjects. Bills Chang sponsored count fully, cosponsored bills a quarter. Shown with at least 20 policy statements and 5 bills.
Ideology
Chang sits at −1.23 on the CSIP scale, left of 0, more liberal than 97% of State Senate Democrats. Approximate range −1.29 to −1.16, which understates the uncertainty; from 656 donors and no roll calls.
CSIP, the lab’s Common-Space Ideal Point, places every official, federal and state, on one left–right scale, estimated from roll-call votes, who gives money to whom, and the text of statements. Negative is left of 0, positive right of it; 0 is the centre of the scale, not a definition of “moderate”.
−1.23
CSIP ideal point
Approximate range −1.29 to −1.16. From 656 donors and no roll calls. Run 2026-10-25T215945. Blue is left of 0, red right of 0.
The outlined dot is Chang, coloured by its side of 0 (blue left, red right, deeper further out); the bar through it is the approximate range (±1.96 standard errors), which understates the uncertainty; the grey ticks below are the medians of Chang’s groups. 0 is the centre of the scale, not “moderate” by definition.
What this means, and what it doesn’t
The score summarises how an official votes, who funds them and how they talk, relative to everyone else. It does not grade positions as good or bad, and two officials with the same score can differ on any single issue. The ± range is 1.96 standard errors, which understates the uncertainty: scores whose ranges overlap may not differ at all, and even scores whose ranges do not overlap may not differ.
Where Chang sits
Every official in each group, on the CSIP scale
State Senate Democrats825 · median −0.7
Hawaii Senate25 · median −0.7
All State Senate1,909 · median +0.1
Show the data
| Group | Officials | Median | Share left of Chang |
|---|---|---|---|
| State Senate Democrats | 825 | −0.7 | 3% |
| Hawaii Senate | 25 | −0.7 | 0% |
| All State Senate | 1909 | +0.1 | 1% |
How to read the strips
Every small dot is one official on the same scale, coloured by its own score (blue left of 0, red right of 0, grey at 0, deeper further out) and spread up and down only to keep dots apart; where a group is crowded the dots overlap, and darker patches mean more officials. The large outlined dot is Chang in every row; the rule is each group’s median. The table under the figure gives the share of each group to the left of Chang.
Money
Chang reported $68.7K in the 2026 cycle, above the State Senate Democrats’ median of $55.5K. State filings report no small-donation total. 92% of itemized individual money came from Hawaii.
Campaign money from state campaign-finance filings, as aggregates only; the lab never shows an individual donor. Campaigns must report by name every person who gives them more than $200 in a cycle (“itemized”); smaller gifts are reported only as one total (“unitemized”), so donor counts, donor states, employers and occupations describe itemized donors only. State filings report no small-gift total, so these figures add up itemized contributions only.
$68.7K
itemized in the 2026 cycle
From 53 itemized donors. State Senate Democrats’ median $55.5K.
Where the money came from
Each square is one percent of itemized contributions in the 2026 cycle.
- Individuals (itemized)65 squares$44.3K
- PACs and other committees35 squares$24.4K
Show the data
| Source | Amount | Share |
|---|---|---|
| Individuals (itemized) | $44.3K | 64.5% |
| PACs and other committees | $24.4K | 35.5% |
What this means, and what it doesn’t
Totals say how much a campaign took in, not how it was spent, and they rise sharply when a seat is up for election; compare cycles with that in mind.
Receipts by cycle
Itemized contributions, 2020 to 2026. The grey line is the State Senate Democrats’ median for 2026.
Show the data
| Item | Value |
|---|---|
| 2020 cycle | $111.7K |
| 2022 cycle | $51.9K |
| 2024 cycle | $62.5K |
| 2026 cycle | $68.7K |
Top donor states, 2026
Itemized contributions by donor state. 92% came from Hawaii.
- Hawaii$40.7K
- California$3,000
- Texas$500
- Delaware$125
Small donations
State filings report no total for small gifts.
Not reported in state filings.
Where itemized donors live
Individuals who gave more than $200, by donor state
Show the data
| State | Value |
|---|---|
| Hawaii | $40.7K |
| California | $3,000 |
| Texas | $500 |
| Delaware | $125 |
| Alaska | No data |
| Maine | No data |
| Vermont | No data |
| New Hampshire | No data |
| Washington | No data |
| Idaho | No data |
| Montana | No data |
| North Dakota | No data |
| Minnesota | No data |
| Illinois | No data |
| Wisconsin | No data |
| Michigan | No data |
| New York | No data |
| Rhode Island | No data |
| Massachusetts | No data |
| Oregon | No data |
| Nevada | No data |
| Wyoming | No data |
| South Dakota | No data |
| Iowa | No data |
| Indiana | No data |
| Ohio | No data |
| Pennsylvania | No data |
| New Jersey | No data |
| Connecticut | No data |
| Utah | No data |
| Colorado | No data |
| Nebraska | No data |
| Missouri | No data |
| Kentucky | No data |
| West Virginia | No data |
| Virginia | No data |
| Maryland | No data |
| Arizona | No data |
| New Mexico | No data |
| Kansas | No data |
| Arkansas | No data |
| Tennessee | No data |
| North Carolina | No data |
| South Carolina | No data |
| District of Columbia | No data |
| Oklahoma | No data |
| Louisiana | No data |
| Mississippi | No data |
| Alabama | No data |
| Georgia | No data |
| Florida | No data |
53
itemized donors
People who gave more than $200
92%
from Hawaii
Of itemized individual dollars
How to read the map
Every state the same size. Darker states gave more itemized money; hatched states had no itemized donors; the outlined state is Hawaii. Gifts of $200 or less carry no address and are left out.
Top donor occupations
Itemized, 2026 cycle.
- ATTORNEY$4,000 6 donors
- RETIRED$3,950 4 donors
- MANAGER$3,000 3 donors
Top committees
PAC and committee contributions, 2026 cycle.
- FRIENDS OF JARRETT KEOHOKALOLE$4,000
- ILWU LOCAL 142 HAWAII POLITICAL ACTION COMMITTEE$4,000
- HAWAII REALTORS POLITICAL ACTION COMMITTEE$2,750
- CASTLE & COOKE INC. LEGISLATIVE COMMITTEE-HI$2,000
- IBEW LOCAL 1260 VOLUNTARY POLITICAL FUND$2,000
Top employers
Itemized, 2026 cycle.
- HIGHRIDGE COSTA$3,000 3 donors
Aggregates only. Individual donors are never named.
Votes
Chang missed 40% of 10 roll calls. Hawaii Senate average attendance 99.5%.
10 roll calls Chang could vote on in the HI 2025–26 session, compared with Hawaii Senate. Attendance is the share with a vote recorded. Party unity counts only party-line roll calls, where most Democrats and most Republicans voted on opposite sides, and is the share of those where Chang voted with the majority of Hawaii Senate Democrats. The count of votes against that majority covers every roll call where the majority took a side.
10
roll calls
No party-line votes. Attendance 60%, ranks 13th of 13.
Attendance and party unity
The thick bar is Chang, the thin grey bar the Hawaii Senate average. The scale runs from 60% to 100% to show the gap.
60%100%
Chang
Peers
Rank
Attendance
60%
99.5%
13th of 13
Party unity
—
—
Show the data
| Row | Chang | Peers |
|---|---|---|
| Attendance | 60% | 99.5% |
| Party unity | — | — |
What this means, and what it doesn’t
A missed roll call has many causes, from illness to campaigning, so attendance says how often, not why. A vote against the majority can be on procedure as well as substance, and many fall on roll calls where both parties split; the key votes below show the ones that decided something.
Key votes
Roll calls the lab flags as key
No votes in this window met the key-vote rules.
How key votes are chosen
Automatically: a final-passage or override vote where the measure became law, the margin was under 10%, at least 10% of either party voted against its majority, or the bill was among the 25 most mentioned in officials’ own statements. Up to 25 per window, most recent first; five are shown. Bars show yea on the left, nay on the right, per party.
Votes most often with
The colleagues who voted the same way as Chang most often
Not enough shared votes.
Votes least often with
The colleagues who voted the same way as Chang least often
Not enough shared votes.
How to read the agreement lists
The share of roll calls both members voted on where they voted the same way, among colleagues sharing at least 50 such votes. High agreement with members of the other party is common on routine votes.
Full voting record
HI 2025–26 session, loads in batches
Legislation
Chang sponsored 2,308 bills in the HI 2025–26 session; 125 became law. Cosponsored 34 others. 47% of sponsored bills got out of committee (Hawaii Senate average 50%; 17th of 24).
The bills Chang sponsored and cosponsored in the HI 2025–26 session, from the official bill records, and how far each got: introduced, referred to or acted on in committee, passed one chamber, passed both, became law. A bill counts at every stage it has reached, so the blocks only shrink to the right.
Bills by stage
Counts for the HI 2025–26 session
Show the data
| Role | Introduced | In committee | Passed one chamber | Passed both chambers | Became law |
|---|---|---|---|---|---|
| Sponsored | 2,308 | 2,308 | 927 | 460 | 125 |
| Cosponsored | 34 | 34 | 17 | 12 | 2 |
How to read the blocks
Ink blocks are bills Chang sponsored, ochre blocks bills Chang cosponsored; the figure above each block is the count. Most bills in any legislature never leave committee, so a short staircase is the norm.
47%
of sponsored bills got out of committee
Hawaii Senate average 50%. Ranks 17th of 24.
Cosponsorship partners
Colleagues who most often cosponsor with Chang, by number of shared bills in the HI 2025–26 session; the square shows their party.
- Joy San Buenaventura1,034
- Troy Hashimoto981
- Angus McKelvey897
- Tim Richards852
- Kurt Fevella844
What this means, and what it doesn’t
The score is the share of sponsored bills that a committee reported out or a chamber passed, with at least 5 sponsored bills, ranked within the peer group. Members of the majority party and committee chairs move more bills, so the rank reflects position in the chamber as well as effort. Cosponsorship partners are the five colleagues who share the most of Chang’s bills; the toggle shows more.
Bill list
Sponsored and cosponsored, with status and policy area; five are shown, the toggle shows the rest
| SR 178 | REQUESTING THE DEPARTMENT OF HEALTH, DEPARTMENT OF HUMAN SERVICES, AND DEPARTMENT OF EDUCATION TO DEVELOP A COORDINATED PLAN TO REDUCE THE IMPACT OF FETAL ALCOHOL SPECTRUM DISORDERS. | Sponsor | Mar 16, 2026 | In committee | Health(model) | Referred to HHS/EDU. Mar 19, 2026 |
| SCR 94 | ENDORSING WAIKIKI AS A WORLD SURFING RESERVE. | Sponsor | Mar 16, 2026 | Passed both chambers | Unassigned | Certified copies of resolutions sent, 05-12-26. May 12, 2026 |
| SR 77 | REQUESTING ALL STATE AGENCIES TO UTILIZE BI-LEVEL LIGHTING SYSTEMS IN ALL STATE BUILDINGS AND FACILITIES, THE DESIGN OF WHICH IS FINALIZED ON OR AFTER JANUARY 1, 2028. | Sponsor | Mar 16, 2026 | Passed one chamber | Energy(model) | Certified copies of resolutions sent, 05-12-26. May 12, 2026 |
| SR 113 | REQUESTING THE DEPARTMENT OF AGRICULTURE AND BIOSECURITY AND DEPARTMENT OF LAW ENFORCEMENT TO INCREASE PENALTIES AND ENFORCEMENT MECHANISMS FOR DOG ATTACKS ON LIVESTOCK. | Sponsor | Mar 16, 2026 | In committee | Agriculture and Food(model) | Referred to AEN/PSM. Mar 19, 2026 |
| SR 101 | REQUESTING THE DEPARTMENT OF HOUSING AND LAND MANAGEMENT OF THE CITY AND COUNTY OF HONOLULU TO DEVELOP A COUNTYWIDE HOUSING PATTERN BOOK TO STREAMLINE PERMIT APPROVALS AND REDUCE REGULATORY BARRIERS FOR SAFE, AFFORDABLE, AND WELL-DESIGNED HOMES. | Sponsor | Mar 16, 2026 | Passed one chamber | Housing and Community Development(model) | Certified copies of resolutions sent, 05-12-26. May 12, 2026 |
Online
A site that changed 1 time in 90 days. Chang’s official site had 1 page when last crawled on Sep 1, 2026.
The lab crawls Chang’s official and campaign websites (weekly for Congress, monthly for the states) and keeps every page’s text, so changes can be read later; it also records social accounts and describes the images Chang posts with a vision model.
Websites
Official site
www.capitol.hawaii.gov/legislature/memberpage.aspx?member=157&year=2017Last crawled Sep 1, 2026
1pages crawled
1changes in 90 days
- 30
pages crawled
27changes in 90 days
How to read the counts
“Pages crawled” is the size of the site at the last crawl; “changes in 90 days” counts pages whose text differed from the previous crawl. Sites change for many routine reasons, such as new press releases, so a high count is not by itself notable.
Images Chang posted
No images from Chang’s posts have been described.
About the descriptions
Written by a vision model and may be wrong. Open the post to see the original. Five are shown; the toggle shows the rest.
State floor speeches
15 turns, Mar 3, 2020 to Apr 30, 2025
Floor speeches
Transcribed from chamber video or captions by the lab. They are not classified for kinds of statement.
Methods and sources
How this page was made. Profile built Oct 28, 2026; site last built Oct 28, 2026. Every number on the page is computed from the sources listed here, by the same rules for every official; nothing is written by hand.
The headline
The headline is the measure on which Chang is most unusual among State Senate Democrats, chosen by one rule for every official of every party. The candidates are the rate of each kind of statement and of extreme labels, party unity, attendance, the small-donor share, the in-state share and total receipts of the latest cycle, the share of sponsored bills that got out of committee, talk-to-bills agreement, and the CSIP score’s place within the party group.
The builder ranks every measure within the same peer group the page compares it with, and writes each official’s percentile: the share of the group below the value, ties counting half, so 0.5 is the median and a value shared by many never ranks first. A measure’s score is how far that percentile sits from the median (0 in the middle, 0.5 at either end), and the highest score leads if it is above 0.25, outside the middle half of the group. Scores within 0.03 of each other are a tie, broken in this order: votes against the party majority, kinds of statement, ideology, roll calls missed, small donations, in-state money, legislation, agreement, extreme labels, receipts.
A measure counts only with at least 200 statements, 100 roll calls or 5 sponsored bills behind it and a peer group of at least 10. Two things never lead: high party unity or attendance, where most members sit, and a high rate of a kind of statement backed by fewer than 10 statements. When nothing qualifies, the headline is a neutral fact instead: statements classified, or years in office. The wording always names the direction (“in the bottom 5% of …”, “votes against the majority of …”).
Identifiers
- Profile id
- d6359473-e0b5-4040-8d7f-9d95a0518650
- Open States
- ocd-person/d6359473-e0b5-4040-8d7f-9d95a0518650
Bulk files with the same data for every official:
Freshness of this profile’s data
- What politicians sayOct 27, 20261 days agoMembers of Congress: tweets since 2017; press releases, newsletters and floor speeches since 2022. State legislators: tweets since May 2024. Governors: tweets since January 2025. Labels from an LLM classifier.
- Roll calls (state)Oct 21, 20267 days agoOpen States floor roll calls, 2019 on (none for AK and MO).
- Bills (state)Oct 28, 2026todayOpen States bills and sponsors, 2019 on; stages from roll calls.
- Campaign money (state)no dateState campaign-finance agencies, 2019 on; aggregates only.
- Ideology (CSIP)Oct 25, 20263 days agoCommon-Space Ideal Points (CSIP), current validated run.
- WebsitesOct 25, 20263 days agoWeekly crawls of federal sites since April 2026; monthly state crawls since June 2026.
- State floor speechesOct 21, 20267 days agoState floor speech turns (published transcripts and captions), 2020 on; not LLM-classified.
How the measures work
- What politicians say
- Every post, press release, newsletter and congressional floor speech is read by a language model and marked for personal attacks, policy criticism, policy discussion, bipartisanship and credit claiming. A statement can be in several. Rates are shares of statements in the window. The methods paper.
- Extreme labels
- A statement has an extreme label when it calls a politician, party or policy “far left”, “far right”, “MAGA”, “extreme”, “radical”, “woke”, “fascist” or “communist”. The target is who the label is aimed at.
- Policy areas and agreement
- Statements and bills share 25 Congressional Research Service policy areas (state bills are assigned by a language model). Agreement is 1 − Jensen–Shannon distance between the two distributions, shown with at least 20 policy statements and 5 bills.
- Ideology (CSIP)
- Common-Space Ideal Points place federal and state officials on one left–right scale, estimated jointly from roll calls, campaign contributions and the text of statements, with standard errors. Only a validated run is shown. The methods paper.
- Campaign money
- Campaigns report by name only donors who give more than $200 in a cycle; smaller gifts are reported as one total. Federal totals and the small-donor share come from the FEC’s financial summaries of each campaign committee. Donor counts, donor states, employers and occupations cover itemized donors only. State filings report no small-gift totals, so state figures are itemized only. The lab never publishes donor names.
- Votes and key votes
- Attendance, party unity and the count of votes against the majority of an official’s own party come from roll calls (Voteview for Congress, chamber records for the states). Key votes are final-passage or override votes that became law, were decided by under 10%, saw at least 10% of a party vote against its majority, or concerned one of the 25 bills most mentioned in officials’ statements.
- Peers and ranks
- Comparisons are with the same party, chamber and level (for example House Democrats); state legislators are compared within their state and chamber for votes. Every comparison names its group. Rank 1 is the highest value in the group; tied values share the average of their ranks, and the percentile is the share of the group below the value, ties counting half.
America’s Political Pulse · Polarization Research Lab